Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

Grant, Richard W.; McCloskey, Jodi K; Hatfield, Meghan; Uratsu, Connie S.; Ralston, James D.; Bayliss, Elizabeth A.; Kennedy, Chris J.

doi:10.1001/jamanetworkopen.2020.29068

Cited by 89 publications

(76 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We recruited three groups of patients representing three distinct profiles of complex care needs identified using methods in our prior work and indicated by our clinical stakeholders as high priority for further study. 25,26 These three complex patient profiles were defined by the following key characteristics: Group A ("obesity, opioid prescription, and low-resourced neighborhood"), Group B ("older, high medical morbidity, emergency department, and hospital use"), and Group C ("older, mental and physical health concerns, and low-resourced neighborhood") (Appendix A-group profile criteria used to inform the purposive sampling). 25 Table 1 describes the characteristics of each group.…”

Section: Participantsmentioning

confidence: 99%

“…25,26 These three complex patient profiles were defined by the following key characteristics: Group A ("obesity, opioid prescription, and low-resourced neighborhood"), Group B ("older, high medical morbidity, emergency department, and hospital use"), and Group C ("older, mental and physical health concerns, and low-resourced neighborhood") (Appendix A-group profile criteria used to inform the purposive sampling). 25 Table 1 describes the characteristics of each group. Recruiters mailed invitation letters to eligible patients then called 4 days later to follow-up on participation.…”

Section: Participantsmentioning

confidence: 99%

See 1 more Smart Citation

Aligning care with the personal values of patients with complex care needs

Tuzzio

Berry

Gleason

et al. 2021

Health Services Research

Self Cite

View full text Add to dashboard Cite

Objective To identify opportunities to align care with the personal values of patients from three distinct groups with complex medical, behavioral, and social needs. Data Sources/Study Setting Between June and August 2019, we conducted semi‐structured interviews with individuals with complex care needs in two integrated health care delivery systems. Study Design Qualitative study using semi‐structured interviews. Data Collection Methods We interviewed three groups of patients at Kaiser Permanente Washington and Kaiser Permanente Colorado representing three distinct profiles of complex care needs: Group A (“obesity, opioid prescription, and low‐resourced neighborhood”), Group B (“older, high medical morbidity, emergency department, and hospital use”), and Group C (“older, mental and physical health concerns, and low‐resourced neighborhood”). These profiles were identified based on prior work and prioritized by internal primary care stakeholders. Interview transcripts were analyzed using thematic analysis. Principal Findings Twenty‐four patients participated; eight from each complex needs profile. Mean age across groups was 71 (range 48–86) years. We identified five themes common across the three groups that captured patients' views regarding values‐aligned care. These themes focused on the importance of care teams exploring and acknowledging a patient's values, providing access to nonphysician providers who have different perspectives on care delivery, offering values‐aligned mental health care, ensuring connection to community‐based resources that support values and address needs, and providing care that supports the patient plus their family and caregivers. Conclusions Our results suggest several opportunities to improve how care is delivered to patients with different complex medical, behavioral, and social needs. Future research is needed to better understand how to incorporate these opportunities into health care.

show abstract

Section: Participantsmentioning

confidence: 99%

Section: Participantsmentioning

confidence: 99%

Aligning care with the personal values of patients with complex care needs

Tuzzio

Berry

Gleason

et al. 2021

Health Services Research

Self Cite

View full text Add to dashboard Cite

show abstract

“…21 However, it has been suggested that the controlled RCT framework may not be the best vehicle for studying the impact of interventions targeting HCU; and realworld, population-based surveillance for upstream identification and intervention may yield better results. 22 In longitudinal analysis, we observed a trajectory of increasing health care costs in the years prior to patients being identified as HCU. The development of learning health care systems that integrate primary and tertiary health data with exogenous factors such as social determinants, and environmental data; and the employment of novel methods, such as machine learning, may offer opportunities for the early identification of potential HCU.…”

Section: J O U R N a L P R E -P R O O Fmentioning

confidence: 94%

Prevalence of Cardiovascular Disease in a Population-Based Cohort of High-Cost Healthcare Services Users

et al. 2022

View full text Add to dashboard Cite

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

show abstract

“…We first carry out an exploratory data analysis, in which we use three sets of variables to cluster patients with K-means clustering methods [ 30 ]. Clustering of patients or locations have been widely used in clinical and healthcare literature to make broad generalizable observations and analysis [ 31 ]. The objective of the clustering analysis is to discover evidence of encounter usage inequality across ZIP codes with varying sociodemographic characteristics, after adjusting for clinical measurements.…”

Section: Study Data and Methodsmentioning

confidence: 99%

“…The advantage of K-means clustering is that it does not require distributional assumptions and can be done in a non-parametric setting. However, K-means depends on the initial starting points, and therefore tends to suffer from lack of stability in some cases [ 31 ]. Therefore, we performed the K-means with 25 random starts and used the most frequent clustering.…”

Section: Study Data and Methodsmentioning

confidence: 99%

Recommending encounters according to the sociodemographic characteristics of patient strata can reduce risks from type 2 diabetes

et al. 2021

View full text Add to dashboard Cite

Objectives Physician encounters with patients with type 2 diabetes act as motivation for self-management and lifestyle adjustments that are indispensable for diabetes treatment. We elucidate the sociodemographic sources of variation in encounter usage and the impact of encounter usage on glucose control, which can be used to recommend encounter usage for different sociodemographic strata of patients to reduce risks from Type 2 diabetes. Data and methods We analyzed data from a multi-facility clinic in the Midwestern United States on 2124 patients with type 2 diabetes, from 95 ZIP codes. A zero-inflated Poisson model was used to estimate the effects of various ZIP-code level sociodemographic variables on the encounter usage. A multinomial logistic regression model was built to estimate the effects of physical and telephonic encounters on patients’ glucose level transitions. Results from the two models were combined in marginal effect analyses. Results and conclusions Conditional on patients’ clinical status, demographics, and insurance status, significant inequality in patient encounters exists across ZIP codes with varying sociodemographic characteristics. One additional physical encounter in a six-month period marginally increases the probability of transition from a diabetic state to a pre-diabetic state by 4.3% and from pre-diabetic to the non-diabetic state by 3.2%. Combined marginal effect analyses illustrate that a ZIP code in the lower quartile of high school graduate percentage among all ZIP codes has 1 fewer physical encounter per six months marginally compared to a ZIP code at the upper quartile, which gives 5.4% average increase in the probability of transitioning from pre-diabetic to diabetic. Our results suggest that policymakers can target particular patient groups who may have inadequate encounters to engage in diabetes care, based on their immediate environmental sociodemographic characteristics, and design programs to increase their encounters to achieve better care outcomes.

show abstract

Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

Abstract: This cohort study uses data clustering methods and clinical stakeholder assessment to identify clinical profiles in a population of medically complex patients.

Cited by 89 publications

References 46 publications

Aligning care with the personal values of patients with complex care needs

Aligning care with the personal values of patients with complex care needs

Prevalence of Cardiovascular Disease in a Population-Based Cohort of High-Cost Healthcare Services Users

Recommending encounters according to the sociodemographic characteristics of patient strata can reduce risks from type 2 diabetes

Contact Info

Product

Resources

About